dsh-doublecheck 0.9.7 → 0.9.8

package/CHANGELOG.md

@@ -2,6 +2,22 @@
 
 All notable changes to dsh-doublecheck are recorded here, newest first.
 
+## [0.9.8] - 2026-09-09
+
+### Fixed
+
+- Adapt the PTC sub-dispatch event vocabulary to the session-format V3 line: the host renamed the durable `tool/code-dispatch` event to `tool/ptc-dispatch` (payload unchanged), so on `0.1.5-alpha.1` the four evidence folds (discipline stage, test evidence, report facts, session projection) silently missed PTC test runs and dispatched `edit`/`write` calls. A shared `ptcSettle()` normalizer now folds the current label and matches the predecessor label structurally, so both the published `0.1.2-rc.1` line and the V3 line stay supported by the unchanged peer band.
+- Point `scripts/scan-sessions.mjs` at the generation-suffixed log filename (`session.v3.jsonl.zstd`, highest `vN` per directory) instead of the hardcoded flat `session.jsonl.zstd`, and require the session project directory as an explicit argument so the tool can never default to the real `~/.dsh` store.
+
+### Changed
+
+- Add the `typecheck:ci` gate (`tsconfig.ci.json`, published-type resolution) and run it in `prepublishOnly`; no behavior change.
+- Extend the CI/compat install probes to a `0.1.2-rc.1` + `0.1.5-alpha.1` matrix instead of replacing the old pin, so both release lines keep an install smoke.
+
+### Docs
+
+- Refresh the five-language README compatibility sections: harness baseline `dsh-v0.1.5-alpha.1`, the `tool/ptc-dispatch` event name, and the `Session.append` / settings-seam facts re-verified on `0.1.5-alpha.1`; no behavior change.
+
 ## [0.9.7] - 2026-09-07
 
 ### Fixed

package/README.es.md

@@ -25,7 +25,7 @@
 
 | Superficie | Estado |
 |---|---|
-| Harness | DeepSeek Harness `dsh-v0.1.3-alpha.1`. Verificado el 2026-09-06 contra el checkout master `dsh-v0.1.3-alpha.1` (cadena completa de puertas + humo de instalación del perfil). |
+| Harness | DeepSeek Harness `dsh-v0.1.5-alpha.1`. Verificado el 2026-09-09 contra el checkout master `dsh-v0.1.5-alpha.1` (cadena completa de puertas + humo de instalación del perfil); la línea publicada `0.1.2-rc.1` sigue soportada. |
 | Node | `^22.19.0 \|\| >=24.0.0` |
 | Plataformas | Todas (host puro; sin código nativo, sin solicitudes de red directas propias) |
 | Modelo | Cualquiera (el guard nunca llama a un modelo; las fases de crítico y revisor se ejecutan como subagentes del harness) |
@@ -208,7 +208,7 @@ El CLI solo serializa el `GateState` ya asentado — nunca vuelve a ejecutar la
 
 ## Permisos y datos
 
-- **Lee**: el registro de sesión (`tool/call` / `tool/result` / `tool/code-dispatch`, fuentes `user/message` inyectadas y los registros de veredicto ajenos `autoReview/*`) solo en proceso; el estado opcional del servicio de modo plan.
+- **Lee**: el registro de sesión (`tool/call` / `tool/result` / `tool/ptc-dispatch`, fuentes `user/message` inyectadas y los registros de veredicto ajenos `autoReview/*`) solo en proceso; el estado opcional del servicio de modo plan. Antes del renombrado de V3 el host registra los subenvíos PTC con la etiqueta predecesora `tool/code-dispatch`; ambas etiquetas se pliegan de forma idéntica.
 - **Escribe**: `doublecheck-spec.md`, `doublecheck-report.md` y `gate-report.md` en el workspace de sesión (rutas configurables) a través de la interfaz `ctx.fs`; los eventos de sesión durables `doublecheck/state` y `doublecheck/gate`.
 - **Llamadas a modelo**: las fases de consistencia y revisión local de la puerta (un subagente cada una por `/gate run`), la revisión adversarial opcional y el flujo de verificación de `doublecheck_report` inician ejecuciones de subagente; nada más llama a un modelo o a la red.
 - **Nunca toca**: credenciales, variables de entorno o cualquier archivo fuera del workspace de sesión. El manifiesto del workshop declara solo `filesystem:read` y `filesystem:write`. Los informes de la puerta llevan solo conteos, ids y veredictos; los secretos reconocidos en los textos del revisor se redactan antes de almacenarlos o mostrarlos.
@@ -225,6 +225,7 @@ El CLI solo serializa el `GateState` ya asentado — nunca vuelve a ejecutar la
 
 - **Escrituras durables.** `/doublecheck on\|off` → `doublecheck/state` y `/gate run` → `doublecheck/gate` necesitan la superficie de append `ignorable` del host (post-rc.6 hasta `0.1.1-rc.2`). En hosts sin esa superficie (rc.6/rc.8 y `0.1.2-alpha.1`, que eliminó el sobre — `0.1.2-rc.1` restaura el campo solo para compatibilidad de lectura de logs almacenados y aún no puede estamparlo), las escrituras se omiten y el interruptor queda en proceso.
 0.1.2-rc.1 (adaptado el 2026-09-02): el sobre de sesión conserva su campo ignorable solo para compatibilidad de lectura de logs almacenados - Session.append aún no puede estamparlo, por lo que el comportamiento de la puerta no cambia.
+0.1.5-alpha.1 (adaptado el 2026-09-09): el formato de sesión V3 renombra el evento durable de subenvío `tool/code-dispatch` a `tool/ptc-dispatch` (carga útil sin cambios; ambas etiquetas se pliegan de forma idéntica). Session.append sigue sin canal `ignorable`, así que las escrituras durables se omiten y el interruptor permanece en proceso - comportamiento sin cambios. El espacio de ajustes `doublecheck.gate` sigue siendo una interfaz débil (véase Limitaciones conocidas).
 - **Interfaces opcionales.** El espacio de ajustes `doublecheck.gate` se registra solo cuando el servicio de ajustes está montado; la línea de modo plan de `/gate status` lee el `ctx.planMode` opcional (muestra `unknown` sin él); la revisión adversarial necesita `ctx.subagents`; la verificación necesita `workflowEngine`.
 - **Degradación local.** `gate.review.engine: auto` degrada al revisor local cuando dsh-auto-review está ausente o no tiene registros de veredicto en esta sesión — el informe nombra la razón en lugar de inventar un veredicto.
 - **La evidencia dsh-eval se basa en archivos.** El motor de evaluación de dsh-auto-review (`dsh-eval`) escribe sus resultados de regresión de prompt / estrés / equidad en un archivo de informe del espacio de trabajo, no en el registro de sesión. `gate.tests.evalReports.enabled` pliega ese archivo (desactivado por defecto; salta cuando está ausente) y los recuentos pliegados viajan en el registro duradero `doublecheck/gate` para que una ejecución asentada siga siendo reproducible.

package/README.hi.md

@@ -25,7 +25,7 @@
 
 | सतह | स्थिति |
 |---|---|
-| Harness | DeepSeek Harness `dsh-v0.1.3-alpha.1`। 2026-09-06 को `dsh-v0.1.3-alpha.1` master checkout के विरुद्ध सत्यापित (पूर्ण द्वार श्रृंखला + प्रोफ़ाइल इंस्टॉल स्मोक)। |
+| Harness | DeepSeek Harness `dsh-v0.1.5-alpha.1`। 2026-09-09 को `dsh-v0.1.5-alpha.1` master checkout के विरुद्ध सत्यापित (पूर्ण द्वार श्रृंखला + प्रोफ़ाइल इंस्टॉल स्मोक); प्रकाशित `0.1.2-rc.1` पिन लाइन अभी भी समर्थित है। |
 | Node | `^22.19.0 \|\| >=24.0.0` |
 | प्लेटफ़ॉर्म | सभी (शुद्ध host; कोई नेटिव कोड नहीं, स्वयं का कोई सीधा नेटवर्क अनुरोध नहीं) |
 | मॉडल | कोई भी (guard स्वयं कभी मॉडल नहीं बुलाता; critic और reviewer चरण harness subagent के रूप में चलते हैं) |
@@ -208,7 +208,7 @@ CLI केवल पहले से स्थापित `GateState` को
 
 ## अनुमतियाँ और डेटा
 
-- **पढ़ता है**: सत्र लॉग (`tool/call` / `tool/result` / `tool/code-dispatch`, इंजेक्ट किए गए `user/message` स्रोत, और बाहरी `autoReview/*` निर्णय रिकॉर्ड) केवल प्रक्रिया के भीतर; वैकल्पिक प्लान-मोड सेवा स्थिति।
+- **पढ़ता है**: सत्र लॉग (`tool/call` / `tool/result` / `tool/ptc-dispatch`, इंजेक्ट किए गए `user/message` स्रोत, और बाहरी `autoReview/*` निर्णय रिकॉर्ड) केवल प्रक्रिया के भीतर; वैकल्पिक प्लान-मोड सेवा स्थिति। V3 नाम-परिवर्तन से पहले का host PTC उप-प्रेषण पूर्ववर्ती लेबल `tool/code-dispatch` से दर्ज करता है; दोनों लेबल समान रूप से फ़ोल्ड होते हैं।
 - **लिखता है**: सत्र कार्यक्षेत्र में `doublecheck-spec.md`, `doublecheck-report.md` और `gate-report.md` (पथ कॉन्फ़िगर करने योग्य) `ctx.fs` इंटरफ़ेस के माध्यम से; टिकाऊ `doublecheck/state` और `doublecheck/gate` सत्र इवेंट।
 - **मॉडल कॉलें**: द्वार के संगति और स्थानीय-समीक्षा चरण (प्रत्येक `/gate run` पर एक-एक subagent), वैकल्पिक प्रतिकूल समीक्षा, और `doublecheck_report` सत्यापन वर्कफ़्लो subagent रन शुरू करते हैं; इसके अलावा कुछ भी मॉडल या नेटवर्क नहीं बुलाता।
 - **कभी नहीं छूता**: क्रेडेंशियल, पर्यावरण चर, या सत्र कार्यक्षेत्र के बाहर कोई फ़ाइल। workshop मेनिफ़ेस्ट केवल `filesystem:read` और `filesystem:write` घोषित करता है। द्वार रिपोर्टें केवल गणना, id और निर्णय रखती हैं; समीक्षक पाठों में पहचाने गए रहस्य भंडारण या प्रदर्शन से पहले संपादित (redacted) कर दिए जाते हैं।
@@ -225,6 +225,7 @@ CLI केवल पहले से स्थापित `GateState` को
 
 - **टिकाऊ लेखन।** `/doublecheck on\|off` → `doublecheck/state` और `/gate run` → `doublecheck/gate` को host की `ignorable` append सतह (rc.6 के बाद से `0.1.1-rc.2` तक) चाहिए। उस सतह के बिना hosts (rc.6/rc.8 और `0.1.2-alpha.1`, जिसने envelope हटा दिया — `0.1.2-rc.1` केवल संग्रहीत-लॉग पठन संगतता के लिए फ़ील्ड बहाल करता है और अभी भी स्टैम्प नहीं कर सकता) पर लेखन छोड़ दिया जाता है और स्विच प्रक्रिया-स्थानीय रहता है।
 0.1.2-rc.1 (2026-09-02 को अनुकूलित): सत्र लिफ़ाफ़ा अपना ignorable फ़ील्ड केवल संग्रहीत-लॉग पठन संगतता के लिए रखता है - Session.append अभी भी इसे स्टैम्प नहीं कर सकता, इसलिए गेट व्यवहार अपरिवर्तित है।
+0.1.5-alpha.1 (2026-09-09 को अनुकूलित): सत्र प्रारूप V3 टिकाऊ उप-प्रेषण इवेंट `tool/code-dispatch` का नाम `tool/ptc-dispatch` करता है (पेलोड अपरिवर्तित; दोनों लेबल समान रूप से फ़ोल्ड होते हैं)। Session.append में अभी भी `ignorable` चैनल नहीं है, इसलिए टिकाऊ लेखन छोड़े जाते हैं और स्विच प्रक्रिया-स्थानीय रहता है - व्यवहार अपरिवर्तित। `doublecheck.gate` सेटिंग्स नेमस्पेस अभी भी एक कमज़ोर इंटरफ़ेस है (ज्ञात सीमाएँ देखें)।
 - **वैकल्पिक इंटरफ़ेस।** `doublecheck.gate` सेटिंग्स नेमस्पेस केवल तब पंजीकृत होता है जब सेटिंग्स सेवा माउंट हो; `/gate status` की प्लान-मोड पंक्ति वैकल्पिक `ctx.planMode` पढ़ती है (इसके बिना `unknown` दिखाती है); प्रतिकूल समीक्षा को `ctx.subagents` चाहिए; सत्यापन को `workflowEngine` चाहिए।
 - **स्थानीय अवनति।** जब dsh-auto-review अनुपस्थित हो या इस सत्र में उसके कोई निर्णय रिकॉर्ड न हों, तो `gate.review.engine: auto` स्थानीय समीक्षक पर घट जाता है — रिपोर्ट निर्णय गढ़ने के बजाय कारण बताती है।
 - **dsh-eval साक्ष्य फ़ाइल-आधारित है।** dsh-auto-review का मूल्यांकन इंजन (`dsh-eval`) अपने prompt-regression / stress / fairness परिणाम कार्यक्षेत्र रिपोर्ट फ़ाइल में लिखता है, सत्र लॉग में नहीं। `gate.tests.evalReports.enabled` उस फ़ाइल को मोड़ता है (डिफ़ॉल्ट रूप से बंद; अनुपस्थित होने पर छोड़ देता है) और मोड़ी गई गणनाएँ टिकाऊ `doublecheck/gate` रिकॉर्ड पर चलती हैं, ताकि तयशुदा रन फिर भी रीप्ले हो सके।

package/README.md

@@ -26,7 +26,7 @@
 
 | Surface | Status |
 |---|---|
-| Harness | DeepSeek Harness `dsh-v0.1.3-alpha.1`. Verified 2026-09-06 against the `dsh-v0.1.3-alpha.1` master checkout (full gate chain + profile install smoke). |
+| Harness | DeepSeek Harness `dsh-v0.1.5-alpha.1`. Verified 2026-09-09 against the `dsh-v0.1.5-alpha.1` master checkout (full gate chain + profile install smoke); the published `0.1.2-rc.1` pin line stays supported. |
 | Node | `^22.19.0 \|\| >=24.0.0` |
 | Platforms | All (pure host; no native code, no direct network requests of its own) |
 | Model | Any (the guard itself never calls a model; the critic and reviewer phases run as harness subagents) |
@@ -209,7 +209,7 @@ The CLI only serializes the already-settled `GateState` — it never re-runs the
 
 ## Permissions & data
 
-- **Reads**: the session log (`tool/call` / `tool/result` / `tool/code-dispatch`, injected `user/message` sources, and the foreign `autoReview/*` verdict records) in-process only; the optional plan-mode service state.
+- **Reads**: the session log (`tool/call` / `tool/result` / `tool/ptc-dispatch`, injected `user/message` sources, and the foreign `autoReview/*` verdict records) in-process only; the optional plan-mode service state. PTC sub-dispatches carry the predecessor `tool/code-dispatch` label on hosts before the V3 rename; both labels fold identically.
 - **Writes**: `doublecheck-spec.md`, `doublecheck-report.md`, `gate-report.md`, and `gate-report.json` in the session workspace (paths configurable) through the `ctx.fs` seam; the durable `doublecheck/state` and `doublecheck/gate` session events.
 - **Model calls**: the gate's consistency and local-review phases (one subagent each per `/gate run`), the optional adversary review, and the `doublecheck_report` verification workflow start subagent runs; nothing else calls a model or the network.
 - **Never touched**: credentials, environment variables, or any file outside the session workspace. The workshop manifest declares `filesystem:read` and `filesystem:write` only. Gate reports carry counts, ids, and verdicts only; recognized secrets in reviewer texts are redacted before storage or display.
@@ -226,6 +226,7 @@ The CLI only serializes the already-settled `GateState` — it never re-runs the
 
 - **Durable writes.** `/doublecheck on\|off` → `doublecheck/state` and `/gate run` → `doublecheck/gate` ride the host's `ignorable` append surface (post-rc.6 through `0.1.1-rc.2`). On hosts without that surface (rc.6/rc.8, and `0.1.2-alpha.1`, which removed the envelope — `0.1.2-rc.1` restores the field for stored-log read compatibility only and still cannot stamp it), the writes are skipped and the switch stays process-local.
 0.1.2-rc.1 (adapted 2026-09-02): the session envelope keeps its ignorable field for stored-log read compatibility only - Session.append still cannot stamp it, so audit-gate behavior is unchanged.
+0.1.5-alpha.1 (adapted 2026-09-09): session format V3 renames the durable sub-dispatch event `tool/code-dispatch` to `tool/ptc-dispatch` (payload unchanged; both labels fold identically). Session.append still exposes no `ignorable` channel, so durable writes stay skipped and the switch stays process-local - behavior unchanged. The `doublecheck.gate` settings namespace remains a weak seam (see Known limitations).
 - **Optional seams.** The `doublecheck.gate` settings namespace registers only when the settings service is mounted; the `/gate status` plan-mode line reads the optional `ctx.planMode` (shows `unknown` without it); the adversary review needs `ctx.subagents`; verification needs `workflowEngine`.
 - **Local degrade.** `gate.review.engine: auto` degrades to the local reviewer when dsh-auto-review is absent or has no verdict records this session — the report names the reason instead of inventing a verdict.
 - **dsh-eval evidence is file-based.** The dsh-auto-review eval engine (`dsh-eval`) writes its prompt-regression / stress / fairness results to a workspace report file, not the session log. `gate.tests.evalReports.enabled` folds that file (off by default; skips when absent) and the folded counts ride the durable `doublecheck/gate` record so a settled run still replays.

package/README.pt.md

@@ -25,7 +25,7 @@
 
 | Superfície | Status |
 |---|---|
-| Harness | DeepSeek Harness `dsh-v0.1.3-alpha.1`. Verificado em 2026-09-06 contra o checkout master `dsh-v0.1.3-alpha.1` (cadeia completa de portões + smoke de instalação do perfil). |
+| Harness | DeepSeek Harness `dsh-v0.1.5-alpha.1`. Verificado em 2026-09-09 contra o checkout master `dsh-v0.1.5-alpha.1` (cadeia completa de portões + smoke de instalação do perfil); a linha publicada `0.1.2-rc.1` continua suportada. |
 | Node | `^22.19.0 \|\| >=24.0.0` |
 | Plataformas | Todas (host puro; sem código nativo, sem requisições de rede diretas próprias) |
 | Modelo | Qualquer (o guard nunca chama um modelo; as fases de crítico e revisor rodam como subagentes do harness) |
@@ -208,7 +208,7 @@ O CLI apenas serializa o `GateState` já assentado — nunca reexecuta a porta d
 
 ## Permissões e dados
 
-- **Lê**: o registro de sessão (`tool/call` / `tool/result` / `tool/code-dispatch`, fontes `user/message` injetadas e os registros de veredicto alheios `autoReview/*`) somente em processo; o estado opcional do serviço de modo plano.
+- **Lê**: o registro de sessão (`tool/call` / `tool/result` / `tool/ptc-dispatch`, fontes `user/message` injetadas e os registros de veredicto alheios `autoReview/*`) somente em processo; o estado opcional do serviço de modo plano. Antes da renomeação da V3, o host registra os subenvios PTC com o rótulo predecessor `tool/code-dispatch`; os dois rótulos são dobrados de forma idêntica.
 - **Escreve**: `doublecheck-spec.md`, `doublecheck-report.md` e `gate-report.md` no workspace da sessão (caminhos configuráveis) por meio da interface `ctx.fs`; os eventos de sessão duráveis `doublecheck/state` e `doublecheck/gate`.
 - **Chamadas a modelo**: as fases de consistência e revisão local do portão (um subagente cada por `/gate run`), a revisão adversarial opcional e o fluxo de verificação de `doublecheck_report` iniciam execuções de subagente; nada mais chama um modelo ou a rede.
 - **Nunca toca**: credenciais, variáveis de ambiente ou qualquer arquivo fora do workspace da sessão. O manifesto do workshop declara apenas `filesystem:read` e `filesystem:write`. Os relatórios do portão carregam apenas contagens, ids e veredictos; segredos reconhecidos nos textos do revisor são redigidos antes de armazenar ou exibir.
@@ -225,6 +225,7 @@ O CLI apenas serializa o `GateState` já assentado — nunca reexecuta a porta d
 
 - **Escritas duráveis.** `/doublecheck on\|off` → `doublecheck/state` e `/gate run` → `doublecheck/gate` precisam da superfície de append `ignorable` do host (pós-rc.6 até `0.1.1-rc.2`). Em hosts sem essa superfície (rc.6/rc.8 e `0.1.2-alpha.1`, que removeu o envelope — `0.1.2-rc.1` restaura o campo apenas para compatibilidade de leitura de logs armazenados e ainda não consegue estampá-lo), as escritas são omitidas e o interruptor permanece em processo.
 0.1.2-rc.1 (adaptado em 2026-09-02): o envelope de sessão mantém seu campo ignorable apenas para compatibilidade de leitura de logs armazenados - o Session.append ainda não consegue estampá-lo, então o comportamento da porta não muda.
+0.1.5-alpha.1 (adaptado em 2026-09-09): o formato de sessão V3 renomeia o evento durável de subenvio `tool/code-dispatch` para `tool/ptc-dispatch` (carga útil inalterada; os dois rótulos são dobrados de forma idêntica). O Session.append continua sem canal `ignorable`, então as escritas duráveis são omitidas e o interruptor permanece em processo - comportamento inalterado. O namespace de configurações `doublecheck.gate` continua sendo uma interface fraca (ver Limitações conhecidas).
 - **Interfaces opcionais.** O namespace de configurações `doublecheck.gate` é registrado apenas quando o serviço de configurações está montado; a linha de modo plano de `/gate status` lê o `ctx.planMode` opcional (mostra `unknown` sem ele); a revisão adversarial precisa de `ctx.subagents`; a verificação precisa de `workflowEngine`.
 - **Degradação local.** `gate.review.engine: auto` degrada para o revisor local quando o dsh-auto-review está ausente ou não tem registros de veredicto nesta sessão — o relatório nomeia a razão em vez de inventar um veredicto.
 - **A evidência dsh-eval é baseada em arquivos.** O motor de avaliação do dsh-auto-review (`dsh-eval`) grava seus resultados de regressão de prompt / estresse / equidade em um arquivo de relatório do espaço de trabalho, não no registro de sessão. `gate.tests.evalReports.enabled` dobra esse arquivo (desativado por padrão; pula quando ausente) e as contagens dobradas viajam no registro durável `doublecheck/gate` para que uma execução assentada ainda seja reproduzível.

package/README.zh.md

@@ -25,7 +25,7 @@
 
 | 方面 | 状态 |
 |---|---|
-| Harness | DeepSeek Harness `dsh-v0.1.3-alpha.1`。已于 2026-09-06 对照 `dsh-v0.1.3-alpha.1` master checkout 核验（完整门禁链 + profile 安装冒烟）。 |
+| Harness | DeepSeek Harness `dsh-v0.1.5-alpha.1`。已于 2026-09-09 对照 `dsh-v0.1.5-alpha.1` master checkout 核验（完整门禁链 + profile 安装冒烟）；已发布的 `0.1.2-rc.1` 钉号线仍受支持。 |
 | Node | `^22.19.0 \|\| >=24.0.0` |
 | 平台 | 全部（纯宿主；无原生代码，自身无直接网络请求） |
 | 模型 | 任意（守卫本身从不调用模型；评审与批评阶段作为宿主 subagent 运行） |
@@ -208,7 +208,7 @@ CLI 只序列化已定的 `GateState` —— 它从不重新运行四阶段门
 
 ## 权限与数据
 
-- **读取**：仅进程内读取会话日志（`tool/call` / `tool/result` / `tool/code-dispatch`、注入的 `user/message` 来源，以及外部的 `autoReview/*` 裁决记录）；可选的计划模式服务状态。
+- **读取**：仅进程内读取会话日志（`tool/call` / `tool/result` / `tool/ptc-dispatch`、注入的 `user/message` 来源，以及外部的 `autoReview/*` 裁决记录）；可选的计划模式服务状态。V3 改名前的宿主用旧标签 `tool/code-dispatch` 记录 PTC 子调用；两个标签折叠结果完全一致。
 - **写入**：会话工作区中的 `doublecheck-spec.md`、`doublecheck-report.md` 和 `gate-report.md`（路径可配置），通过 `ctx.fs` 接口；持久的 `doublecheck/state` 和 `doublecheck/gate` 会话事件。
 - **模型调用**：门禁的一致性阶段和本地评审阶段（每次 `/gate run` 各一个 subagent）、可选的对抗式评审，以及 `doublecheck_report` 验证工作流会启动 subagent 运行；除此之外不调用模型或网络。
 - **绝不触碰**：凭据、环境变量，或会话工作区之外的任何文件。workshop 清单只声明 `filesystem:read` 和 `filesystem:write`。门禁报告只携带计数、id 和裁决；评审文本中被识别的机密在存储或显示之前会被脱敏。
@@ -225,6 +225,7 @@ CLI 只序列化已定的 `GateState` —— 它从不重新运行四阶段门
 
 - **持久写入。** `/doublecheck on\|off` → `doublecheck/state` 与 `/gate run` → `doublecheck/gate` 需要宿主的 `ignorable` 追加接口（rc.6 之后至 `0.1.1-rc.2`）。在无此接口的宿主上（rc.6/rc.8，以及移除该信封的 `0.1.2-alpha.1`——`0.1.2-rc.1` 仅恢复存量日志读取兼容字段、仍无法盖章），写入被跳过、开关保持进程内状态。
 0.1.2-rc.1（2026-09-02 已适配）：会话信封保留 ignorable 字段但仅用于存量日志读取兼容——Session.append 仍无法盖章，门控行为不变。
+0.1.5-alpha.1（2026-09-09 已适配）：会话格式 V3 把持久子调用事件 `tool/code-dispatch` 改名为 `tool/ptc-dispatch`（载荷不变；两个标签折叠结果一致）。Session.append 仍无 `ignorable` 写入通道，写入继续跳过、开关保持进程内——行为不变。`doublecheck.gate` 设置命名空间仍是弱接口（见「已知限制」）。
 - **可选接口。** `doublecheck.gate` 设置命名空间仅在挂载设置服务时注册；`/gate status` 的计划模式行读取可选的 `ctx.planMode`（没有则显示 `unknown`）；对抗式评审需要 `ctx.subagents`；验证需要 `workflowEngine`。
 - **本地降级。** 当 dsh-auto-review 缺失或本会话没有裁决记录时，`gate.review.engine: auto` 会降级到本地评审者——报告会写明原因，而不是捏造裁决。
 - **dsh-eval 证据基于文件。** dsh-auto-review 评测引擎（`dsh-eval`）把其 prompt 回归 / 压测 / 公平性结果写入工作区报告文件，而非会话日志。`gate.tests.evalReports.enabled` 折叠该文件（默认关闭；缺失时跳过），折叠出的计数随持久 `doublecheck/gate` 记录保存，使已定的运行仍可重放。

package/lib/domain/evidence.d.ts

@@ -2,13 +2,14 @@
  * Test-run evidence classification for the red/green gates.
  *
  * The durable session log is the only source of truth: a shell tool call
- * (`bash` / `pwsh`, including Code Mode sub-dispatches) whose command matches
- * the configured test patterns is a test run, and its rendered result text
+ * (`bash` / `pwsh`, including PTC sub-dispatches) whose command matches the
+ * configured test patterns is a test run, and its rendered result text
  * carries the exit facts (`[exit code: N]`, timeout, signal, sandbox-denial
  * markers). A failing run is red evidence; a passing run is green evidence.
  *
  * @module dsh-doublecheck/domain/evidence
  */
+import type { SessionEvent } from '@deepseek-ai/dsh-session';
 /** Compiled red/green evidence detection knobs. */
 export interface TestRunDetection {
     /** Tool names that can execute shell commands. */
@@ -45,9 +46,41 @@ export interface DetectionConfig {
 export declare function compileDetection(config: DetectionConfig): TestRunDetection;
 /** The detection record with every list empty: test-run evidence is ignored. */
 export declare function emptyDetection(): TestRunDetection;
+/** The rendered result blocks of a settled sub-dispatch (structural: only the fields the evidence folds read). */
+export type PtcContent = ReadonlyArray<{
+    type?: unknown;
+    text?: unknown;
+    content?: unknown;
+}>;
+/** One settled PTC sub-dispatch payload, whatever generation names the event. */
+export interface PtcSettle {
+    /** The dispatched tool name (`bash` / `pwsh` / `edit` / `write` …). */
+    name: string;
+    /** The dispatch arguments: a normalized record on the current line, a JSON string on the predecessor. */
+    arguments: unknown;
+    /** Whether the sub-dispatch failed at the infrastructure level. */
+    isError: boolean;
+    /** The rendered model-facing result blocks. */
+    content: PtcContent;
+}
+/**
+ * Normalize one settled PTC sub-dispatch event, across the event-vocabulary
+ * rename at the session-format V2→V3 edge.
+ *
+ * The current line names the settled sub-dispatch `tool/ptc-dispatch` (the
+ * host's V2→V3 migration renames the predecessor `tool/code-dispatch`, keeping
+ * the payload values), so a stored V2 log arrives under the new label. A host
+ * on the older release line still emits the predecessor label, which is outside
+ * this line's `SessionEventMap`; it is matched structurally instead of as a
+ * switch case, so the fold never depends on which generation wrote the log.
+ *
+ * @param event - one committed session event.
+ * @returns the settled payload, or `undefined` when the event is not a settled sub-dispatch.
+ */
+export declare function ptcSettle(event: SessionEvent): PtcSettle | undefined;
 /**
  * Parse a raw tool-call `arguments` value (a JSON string from the model, or an
- * already-normalized object from a Code Mode sub-dispatch) into a plain record.
+ * already-normalized object from a PTC sub-dispatch) into a plain record.
  * @param raw - the raw arguments value from the durable event.
  * @returns the parsed record, or `undefined` when the string is not valid JSON.
  */

package/lib/domain/evidence.js

@@ -2,8 +2,8 @@
  * Test-run evidence classification for the red/green gates.
  *
  * The durable session log is the only source of truth: a shell tool call
- * (`bash` / `pwsh`, including Code Mode sub-dispatches) whose command matches
- * the configured test patterns is a test run, and its rendered result text
+ * (`bash` / `pwsh`, including PTC sub-dispatches) whose command matches the
+ * configured test patterns is a test run, and its rendered result text
  * carries the exit facts (`[exit code: N]`, timeout, signal, sandbox-denial
  * markers). A failing run is red evidence; a passing run is green evidence.
  *
@@ -56,9 +56,44 @@ export function emptyDetection() {
         testFilePatterns: [],
     };
 }
+/**
+ * Normalize one settled PTC sub-dispatch event, across the event-vocabulary
+ * rename at the session-format V2→V3 edge.
+ *
+ * The current line names the settled sub-dispatch `tool/ptc-dispatch` (the
+ * host's V2→V3 migration renames the predecessor `tool/code-dispatch`, keeping
+ * the payload values), so a stored V2 log arrives under the new label. A host
+ * on the older release line still emits the predecessor label, which is outside
+ * this line's `SessionEventMap`; it is matched structurally instead of as a
+ * switch case, so the fold never depends on which generation wrote the log.
+ *
+ * @param event - one committed session event.
+ * @returns the settled payload, or `undefined` when the event is not a settled sub-dispatch.
+ */
+export function ptcSettle(event) {
+    if (event.type === 'tool/ptc-dispatch') {
+        return {
+            name: event.data.name,
+            arguments: event.data.arguments,
+            isError: event.data.isError,
+            content: event.data.content,
+        };
+    }
+    if (event.type !== 'tool/code-dispatch')
+        return undefined;
+    const legacy = event.data;
+    if (legacy === null || legacy === undefined || typeof legacy.name !== 'string')
+        return undefined;
+    return {
+        name: legacy.name,
+        arguments: legacy.arguments,
+        isError: legacy.isError === true,
+        content: Array.isArray(legacy.content) ? legacy.content : [],
+    };
+}
 /**
  * Parse a raw tool-call `arguments` value (a JSON string from the model, or an
- * already-normalized object from a Code Mode sub-dispatch) into a plain record.
+ * already-normalized object from a PTC sub-dispatch) into a plain record.
  * @param raw - the raw arguments value from the durable event.
  * @returns the parsed record, or `undefined` when the string is not valid JSON.
  */

package/lib/domain/gate.js

@@ -25,7 +25,7 @@
  *
  * @module dsh-doublecheck/domain/gate
  */
-import { isTestCommand, joinTextBlocks, parseRawArguments, shellCommand, testOutcome, } from "./evidence.js";
+import { isTestCommand, joinTextBlocks, parseRawArguments, ptcSettle, shellCommand, testOutcome, } from "./evidence.js";
 import { SPEC_TOOL_NAME } from "./stages.js";
 import { SPEC_FIELD_NAMES } from "./vocabulary.js";
 /** All gate phases in checklist order. */
@@ -110,6 +110,19 @@ export function foldTestEvidence(events, detection, coverageRegex) {
     const evidence = { failed: 0, passed: 0, lastOutcome: 'none', failingAfterGreen: 0, coveragePct: null };
     const pendingTests = new Map();
     for (const event of events) {
+        // A settled PTC sub-dispatch is a test run exactly like a native shell
+        // call; the normalizer accepts both event generations, so the evidence
+        // does not depend on which one wrote the log.
+        const settle = ptcSettle(event);
+        if (settle !== undefined) {
+            const args = parseRawArguments(settle.arguments);
+            const command = shellCommand(settle.name, args, detection);
+            if (command !== undefined && isTestCommand(command, detection)) {
+                const text = joinTextBlocks(settle.content);
+                foldTestEvidenceOutcome(evidence, testOutcome(text, settle.isError), text, coverageRegex);
+            }
+            continue;
+        }
         switch (event.type) {
             case 'tool/call': {
                 const args = parseRawArguments(event.data.arguments);
@@ -127,15 +140,6 @@ export function foldTestEvidence(events, detection, coverageRegex) {
                 foldTestEvidenceOutcome(evidence, testOutcome(text, event.data.error !== undefined), text, coverageRegex);
                 break;
             }
-            case 'tool/code-dispatch': {
-                const args = parseRawArguments(event.data.arguments);
-                const command = shellCommand(event.data.name, args, detection);
-                if (command !== undefined && isTestCommand(command, detection)) {
-                    const text = joinTextBlocks(event.data.content);
-                    foldTestEvidenceOutcome(evidence, testOutcome(text, event.data.isError), text, coverageRegex);
-                }
-                break;
-            }
         }
     }
     return evidence;

package/lib/domain/projection.js

@@ -10,7 +10,7 @@
  *
  * @module dsh-doublecheck/domain/projection
  */
-import { isTestCommand, isTestFilePath, joinTextBlocks, mutationTargetPath, parseRawArguments, shellCommand, testOutcome, } from "./evidence.js";
+import { isTestCommand, isTestFilePath, joinTextBlocks, mutationTargetPath, parseRawArguments, ptcSettle, shellCommand, testOutcome, } from "./evidence.js";
 import { SPEC_TOOL_NAME } from "./stages.js";
 import { countRedChecks } from "./gate.js";
 /** State for the empty log. */
@@ -38,6 +38,22 @@ export function emptyDoublecheckState() {
  *   discipline fact.
  */
 export function applyDoublecheckEvent(state, event, detection) {
+    // A settled PTC sub-dispatch is a discipline fact exactly like a native
+    // call. The normalizer accepts both event generations, so the projection
+    // does not depend on which one wrote the log.
+    const settle = ptcSettle(event);
+    if (settle !== undefined) {
+        const args = parseRawArguments(settle.arguments);
+        const command = shellCommand(settle.name, args, detection);
+        if (command !== undefined && isTestCommand(command, detection)) {
+            const outcome = testOutcome(joinTextBlocks(settle.content), settle.isError);
+            if (outcome === 'fail')
+                return { ...state, color: 'red', stage: 'red' };
+            if (outcome === 'pass')
+                return { ...state, color: 'green', stage: 'green' };
+        }
+        return state;
+    }
     switch (event.type) {
         case 'tool/call': {
             const args = parseRawArguments(event.data.arguments);
@@ -82,18 +98,6 @@ export function applyDoublecheckEvent(state, event, detection) {
             }
             return state;
         }
-        case 'tool/code-dispatch': {
-            const args = parseRawArguments(event.data.arguments);
-            const command = shellCommand(event.data.name, args, detection);
-            if (command !== undefined && isTestCommand(command, detection)) {
-                const outcome = testOutcome(joinTextBlocks(event.data.content), event.data.isError);
-                if (outcome === 'fail')
-                    return { ...state, color: 'red', stage: 'red' };
-                if (outcome === 'pass')
-                    return { ...state, color: 'green', stage: 'green' };
-            }
-            return state;
-        }
         case 'user/message': {
             const source = event.data.source;
             if (source.kind === 'doublecheck-review' && !state.reviewed) {

package/lib/domain/report.js

@@ -11,7 +11,7 @@
  *
  * @module dsh-doublecheck/domain/report
  */
-import { isTestCommand, isTestFilePath, joinTextBlocks, mutationTargetPath, parseRawArguments, shellCommand, testOutcome, } from "./evidence.js";
+import { isTestCommand, isTestFilePath, joinTextBlocks, mutationTargetPath, parseRawArguments, ptcSettle, shellCommand, testOutcome, } from "./evidence.js";
 import { SPEC_TOOL_NAME } from "./stages.js";
 import { VERIFY_DIMENSIONS, } from "./vocabulary.js";
 /** Maximum timeline rows rendered into the report document. */
@@ -32,6 +32,23 @@ export function foldReportFacts(events, detection) {
     /** Spec calls awaiting a result: only a successful pair commits the spec. */
     const pendingSpecs = new Map();
     for (const event of events) {
+        // A settled PTC sub-dispatch runs through the same pre-execute policy
+        // gates, so a dispatched test run or edit is a real delivery fact and must
+        // count exactly like a native call. The normalizer accepts both event
+        // generations, so the fold does not depend on which one wrote the log.
+        const settle = ptcSettle(event);
+        if (settle !== undefined) {
+            const args = parseRawArguments(settle.arguments);
+            const command = shellCommand(settle.name, args, detection);
+            if (command !== undefined && isTestCommand(command, detection)) {
+                foldTestRun(facts, command, testOutcome(joinTextBlocks(settle.content), settle.isError));
+            }
+            const path = mutationTargetPath(settle.name, args, detection);
+            if (path !== undefined && !isTestFilePath(path, detection)) {
+                facts.edits += 1;
+            }
+            continue;
+        }
         switch (event.type) {
             case 'tool/call': {
                 const args = parseRawArguments(event.data.arguments);
@@ -68,21 +85,6 @@ export function foldReportFacts(events, detection) {
                 foldTestRun(facts, command, testOutcome(joinTextBlocks(event.data.message.content), event.data.error !== undefined));
                 break;
             }
-            case 'tool/code-dispatch': {
-                const args = parseRawArguments(event.data.arguments);
-                const command = shellCommand(event.data.name, args, detection);
-                if (command !== undefined && isTestCommand(command, detection)) {
-                    foldTestRun(facts, command, testOutcome(joinTextBlocks(event.data.content), event.data.isError));
-                }
-                // Code Mode dispatches run through the same pre-execute policy gates,
-                // so a dispatched edit is a real implementation edit and must count
-                // toward the report exactly like a native call.
-                const path = mutationTargetPath(event.data.name, args, detection);
-                if (path !== undefined && !isTestFilePath(path, detection)) {
-                    facts.edits += 1;
-                }
-                break;
-            }
             case 'user/message': {
                 const source = event.data.source;
                 if (source.kind !== 'doublecheck-review')

package/lib/domain/stages.js

@@ -14,7 +14,7 @@
  *
  * @module dsh-doublecheck/domain/stages
  */
-import { emptyDetection, isTestCommand, isTestFilePath, joinTextBlocks, mutationTargetPath, parseRawArguments, shellCommand, testOutcome, } from "./evidence.js";
+import { emptyDetection, isTestCommand, isTestFilePath, joinTextBlocks, mutationTargetPath, parseRawArguments, ptcSettle, shellCommand, testOutcome, } from "./evidence.js";
 /** All discipline stages in order. */
 export const DISCIPLINE_STAGES = [
     'grill', 'design', 'red', 'green', 'review', 'verify',
@@ -82,6 +82,24 @@ export function foldDisciplineRange(state, events, start, detection) {
         const event = events[index];
         if (event === undefined)
             continue;
+        // A settled PTC sub-dispatch runs through the same pre-execute policy
+        // gates, so a dispatched test run or edit is a real discipline fact and
+        // must count exactly like a native call. The normalizer accepts both event
+        // generations, so the fold does not depend on which one wrote the log.
+        const settle = ptcSettle(event);
+        if (settle !== undefined) {
+            const args = parseRawArguments(settle.arguments);
+            const command = shellCommand(settle.name, args, detection);
+            if (command !== undefined && isTestCommand(command, detection)) {
+                foldTestOutcome(state, testOutcome(joinTextBlocks(settle.content), settle.isError));
+            }
+            const path = mutationTargetPath(settle.name, args, detection);
+            if (path !== undefined && !isTestFilePath(path, detection)) {
+                state.pendingGreen = true;
+                state.editCount += 1;
+            }
+            continue;
+        }
         switch (event.type) {
             case 'tool/call': {
                 const args = parseRawArguments(event.data.arguments);
@@ -118,22 +136,6 @@ export function foldDisciplineRange(state, events, start, detection) {
                 }
                 break;
             }
-            case 'tool/code-dispatch': {
-                const args = parseRawArguments(event.data.arguments);
-                const command = shellCommand(event.data.name, args, detection);
-                if (command !== undefined && isTestCommand(command, detection)) {
-                    foldTestOutcome(state, testOutcome(joinTextBlocks(event.data.content), event.data.isError));
-                }
-                // Code Mode dispatches run through the same pre-execute policy gates,
-                // so a dispatched edit is a real implementation edit and must count
-                // toward the green gate and the report exactly like a native call.
-                const path = mutationTargetPath(event.data.name, args, detection);
-                if (path !== undefined && !isTestFilePath(path, detection)) {
-                    state.pendingGreen = true;
-                    state.editCount += 1;
-                }
-                break;
-            }
         }
     }
     return state;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-doublecheck",
-  "version": "0.9.7",
+  "version": "0.9.8",
   "description": "Delivery quality gate for DeepSeek Harness: grill the requirements, test the implementation, prove the delivery, and gate the handoff with a deliverable/rework decision.",
   "packageManager": "pnpm@11.7.0",
   "repository": {
@@ -56,8 +56,9 @@
   "scripts": {
     "build": "tsc --noEmitOnError && node scripts/fix-dts.mjs",
     "prepare": "tsc --noEmitOnError && node scripts/fix-dts.mjs",
-    "prepublishOnly": "tsc --noEmitOnError && node scripts/fix-dts.mjs && pnpm run test",
+    "prepublishOnly": "pnpm run typecheck:ci && tsc --noEmitOnError && node scripts/fix-dts.mjs && pnpm run test",
     "typecheck": "tsc --noEmit && tsc -p tsconfig.tests.json",
+    "typecheck:ci": "tsc -p tsconfig.ci.json --noEmit",
     "lint": "eslint src tests",
     "test": "vitest run",
     "test:coverage": "vitest run --coverage",
@@ -110,9 +111,7 @@
       "filesystem:write"
     ],
     "compatibility": {
-      "dshVersions": [
-        "0.1.2-rc.1"
-      ]
+      "dshVersions": ["0.1.2-rc.1","0.1.5-alpha.1"]
     },
     "capability": {
       "id": "doublecheck",
@@ -129,14 +128,14 @@
   },
   "peerDependencies": {
     "@deepseek-ai/cordis": "^4.0.2",
-    "@deepseek-ai/dsh-agent": ">=0.1.2-rc.1 <0.2.0",
-    "@deepseek-ai/dsh-commands": ">=0.1.2-rc.1 <0.2.0",
-    "@deepseek-ai/dsh-invariants": ">=0.1.2-rc.1 <0.2.0",
-    "@deepseek-ai/dsh-llm": ">=0.1.2-rc.1 <0.2.0",
-    "@deepseek-ai/dsh-session": ">=0.1.2-rc.1 <0.2.0",
-    "@deepseek-ai/dsh-session-projection": ">=0.1.2-rc.1 <0.2.0",
-    "@deepseek-ai/dsh-skill": ">=0.1.2-rc.1 <0.2.0",
-    "@deepseek-ai/dsh-tools": ">=0.1.2-rc.1 <0.2.0",
+    "@deepseek-ai/dsh-agent": ">=0.1.2-rc.1 <0.2.0 || >=0.1.5-alpha.1 <0.2.0",
+    "@deepseek-ai/dsh-commands": ">=0.1.2-rc.1 <0.2.0 || >=0.1.5-alpha.1 <0.2.0",
+    "@deepseek-ai/dsh-invariants": ">=0.1.2-rc.1 <0.2.0 || >=0.1.5-alpha.1 <0.2.0",
+    "@deepseek-ai/dsh-llm": ">=0.1.2-rc.1 <0.2.0 || >=0.1.5-alpha.1 <0.2.0",
+    "@deepseek-ai/dsh-session": ">=0.1.2-rc.1 <0.2.0 || >=0.1.5-alpha.1 <0.2.0",
+    "@deepseek-ai/dsh-session-projection": ">=0.1.2-rc.1 <0.2.0 || >=0.1.5-alpha.1 <0.2.0",
+    "@deepseek-ai/dsh-skill": ">=0.1.2-rc.1 <0.2.0 || >=0.1.5-alpha.1 <0.2.0",
+    "@deepseek-ai/dsh-tools": ">=0.1.2-rc.1 <0.2.0 || >=0.1.5-alpha.1 <0.2.0",
     "@deepseek-ai/schemastery": "^3.18.2"
   },
   "dependencies": {
@@ -146,17 +145,17 @@
   "devDependencies": {
     "@deepseek-ai/cordis": "^4.0.2",
     "@deepseek-ai/cordis-plugin-loader": "^1.0.3",
-    "@deepseek-ai/dsh-agent": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-commands": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-fs": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-llm": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-session": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-session-projection": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-skill": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-subagent": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-system-prompt": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-tools": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-workflow": "0.1.2-rc.1",
+    "@deepseek-ai/dsh-agent": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-commands": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-fs": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-llm": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-session": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-session-projection": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-skill": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-subagent": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-system-prompt": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-tools": "0.1.5-alpha.1",
+    "@deepseek-ai/dsh-workflow": "0.1.5-alpha.1",
     "@deepseek-ai/schemastery": "^3.18.2",
     "@eslint/js": "^9.30.0",
     "@types/node": "^22.15.0",